Stable-Isotope Lipidomics for De Novo Lipogenesis: Tracer Selection, Isotopomer Spectral Analysis, and Flux Interpretation

Stable-isotope lipidomics can determine whether a lipid pool is being synthesized, which precursor contributes carbon, and how newly synthesized fatty acids are allocated among lipid classes. These are distinct measurements. A higher M+2 signal after tracer exposure establishes label incorporation, but it does not by itself quantify de novo lipogenesis (DNL) flux. Flux interpretation requires a defined product pool, measured precursor enrichment or an appropriate precursor-dilution model, a labeling interval, and evidence that the isotopologue signal represents the molecular segment being claimed.

The experimental design therefore begins with the biological claim rather than the isotope. A study asking whether glucose supplies carbon to palmitate needs a carbon tracer and positional evidence. A study asking what fraction of palmitate was newly synthesized during an interval may be better served by deuterated water. A study asking whether newly synthesized palmitate was incorporated into triacylglycerol, phosphatidylcholine, or another lipid class requires intact-lipid analysis in addition to, or instead of, hydrolyzed fatty-acid measurement.

De Novo Lipogenesis Endpoints in Stable-Isotope Lipidomics

DNL converts acetyl-CoA-derived two-carbon units into fatty acids through acetyl-CoA carboxylase and fatty acid synthase. Palmitate is the principal direct product of mammalian fatty acid synthase, while longer and unsaturated fatty acids arise through additional elongation and desaturation steps. The synthesized acyl chains may then enter triacylglycerols, phospholipids, cholesteryl esters, sphingolipids, or other complex lipids. A single experiment can observe several of these layers, but each layer supports a different statement.

These endpoints should be specified in the study plan and retained as separate fields in the final report. "Percent labeled palmitate," "fractional palmitate synthesis," and "nanomoles of newly synthesized palmitate per hour" are not interchangeable. The first describes an observed distribution, the second is a model-derived fraction, and the third additionally requires an amount and time basis.

The product definition also matters. Palmitate measured after hydrolysis of total cellular lipid integrates palmitate from multiple lipid classes. Palmitate measured within TG(16:0_18:1_18:1) describes allocation into a selected intact species and can include labeling in the glycerol backbone. A rigorous metabolic flux analysis study should therefore define whether the endpoint is total fatty-acid synthesis, neutral-lipid storage, membrane-lipid remodeling, or pathway-specific carbon contribution before the tracer and analytical platform are selected.

Figure 1. [U-13C6]glucose, [1,2-13C2]acetate, and 2H2O enter the lipogenic network through different biochemical routes and answer different questions. Label can appear in fatty acyl chains, the glycerol backbone, or selected polar headgroups, so the measured molecular segment must match the intended endpoint.

13C-Glucose, 13C-Acetate, and D2O Tracer Selection for DNL

Tracer choice determines the interpretation more strongly than instrument mass accuracy. The useful question is not which tracer produces the largest signal, but which precursor-product relationship matches the biological decision.

[U-13C6]Glucose for Glucose Carbon Contribution

[U-13C6]glucose follows glycolysis to pyruvate, mitochondrial acetyl-CoA, citrate export, and cytosolic acetyl-CoA. Repeated incorporation of labeled two-carbon units can generate a series of even-mass fatty-acid isotopologues. The distribution contains information about glucose-derived acetyl-CoA enrichment and the fraction of fatty acid synthesized during the exposure. It is well suited to experiments evaluating glucose-dependent lipogenesis, ACLY-dependent carbon supply, nutrient rewiring, or the effect of a perturbation on glucose contribution to fatty acids.

The limitation is precursor dilution. Cytosolic acetyl-CoA can also arise from acetate, glutamine-dependent citrate, exogenous fatty-acid turnover, or other sources. Consequently, low palmitate labeling does not necessarily mean low total DNL; it may indicate that unlabeled precursors supplied much of the acetyl-CoA pool. Glucose also labels glycerol-3-phosphate and can therefore place three 13C atoms into a glycerol backbone. In intact triacylglycerol or phospholipid spectra, backbone labeling can overlap with or accompany fatty-acyl labeling. Some headgroups can also become labeled through connected metabolism. Precursor and product fragments are needed when the claim is specifically about acyl-chain synthesis.

[1,2-13C2]Acetate for Acetate-Derived Acetyl-CoA

[1,2-13C2]acetate is converted to acetyl-CoA through acetyl-CoA synthetases, including ACSS2 in the cytosolic/nuclear compartment. Its two-carbon labeling pattern is convenient for ISA because newly synthesized fatty acids acquire regular M+2 increments. This tracer can test acetate utilization, ACSS2 dependence, or lipogenesis in models where acetate is a biologically relevant carbon source.

Acetate is not a neutral substitute for glucose. It enters downstream of glycolysis, pyruvate dehydrogenase, and citrate cleavage, and can therefore bypass regulatory steps that are central to the hypothesis. Added acetate can also change substrate availability if the total concentration exceeds the model's normal exposure. A comparison across treatment groups should maintain the same total acetate concentration and change only the labeled fraction where practical. Interpretation should be framed as acetate contribution and synthesis under the defined culture condition, not as an automatic measure of total physiological DNL.

2H2O for Fractional DNL from Multiple Carbon Sources

Deuterated water labels fatty acids through exchange with body or medium water and through deuterium incorporation into intermediates and reducing equivalents. Because the water pool is shared across multiple nutrient routes, 2H2O can estimate the fraction of fatty acid newly synthesized from all carbon sources during the labeling interval. This feature is useful when the study objective is total fractional DNL rather than attribution to one nutrient.

The method requires measurement of water enrichment and an appropriate effective number of biosynthetically incorporated hydrogens for the analyte and model. Natural 13C abundance can interfere with low-level deuterium measurement at nominal mass resolution; high-resolution GC-MS approaches can resolve 2H from 13C contributions and improve sensitivity. Deuterated water cannot identify whether the carbon came from glucose, acetate, or glutamine. It answers a total synthesis question and may need to be paired with a carbon tracer when substrate origin is also required.

Dual-tracer designs can be informative, but overlapping isotopic envelopes and additional model parameters increase analytical risk. They should be used when the second tracer resolves a defined ambiguity, not merely to create a more complex dataset.

Labeling Time Course and Precursor Enrichment

The labeling interval determines what process is visible. Early time points emphasize precursor entry and nascent product formation. Later time points integrate synthesis, elongation, desaturation, lipid-class transfer, turnover, secretion, and approach to isotopic steady state. A single endpoint cannot reveal whether two groups differ in initial synthesis rate, delayed precursor equilibration, or subsequent product loss.

A practical pilot includes multiple time points spanning the expected response, an unlabeled control processed with every analytical batch, and a time-zero or immediate-quench control where feasible. The goal is to identify an interval in which labeled product increases reproducibly without complete saturation or major loss of viability. For cultured cells, cell count, protein or DNA content, viability, growth phase, and confluence should be recorded because biomass expansion can dominate lipid synthesis. For tissues or in vivo work, feeding state, circadian timing, administration route, water enrichment, and tissue collection time become part of the flux definition.

The labeled nutrient should not unintentionally become the perturbation. Where feasible, cells can be pre-equilibrated with the unlabeled precursor and then switched to an isotopically labeled form at the same total concentration. This minimizes differences in nutrient availability between the pre-label and label periods. The approach has been used in tracer-assisted lipidomics to preserve comparability between dynamic tracing and steady-state lipid profiling. It does not eliminate isotope effects or medium exchange artifacts, but it reduces a common design confound.

Precursor enrichment must be measured at the level relevant to the model. Medium enrichment is an input, not necessarily the intracellular acetyl-CoA enrichment experienced by fatty acid synthase. Rapid precursor turnover, compartmentation, and unlabeled nutrient sources can create substantial dilution. ISA addresses part of this problem by inferring an effective lipogenic acetyl-CoA enrichment from the product isotopologue pattern. Direct measurement of acetyl-CoA may provide complementary evidence, but sample stability, compartment mixing, and analytical coverage should be acknowledged.

Biological replicates remain the unit of biological inference. Repeated injections improve confidence in technical precision but do not replace independent cultures, animals, tissue donors, or experimental preparations. Randomized extraction and injection order, pooled quality-control samples, blanks, and internal standards are required when small changes in isotopologue abundance are compared across a long batch.

Fatty-Acid and Intact-Lipid Isotopologue Readouts

The analytical route should follow the molecular level at which the claim is made. Hydrolysis and derivatization simplify the isotope envelope of individual fatty acids. Intact-lipid analysis preserves lipid-class and molecular-species allocation but introduces multiple label locations and a larger isotopologue search space.

Hydrolyzed fatty-acid analysis is often the clearest route for palmitate ISA. Following saponification or transesterification, fatty-acid methyl esters can be analyzed by GC-MS and modeled without the combinatorial complexity of an intact triacylglycerol. The sample preparation must nevertheless prevent isotopic exchange, incomplete recovery, oxidation, and carryover. A dedicated fatty acid profiling workflow by GC-FID or GC-MS can complement isotope measurements by providing the fatty-acid pool sizes required to translate fractional synthesis into newly synthesized amount.

Intact-lipid LC-HRMS answers a different question. If a TG species shifts by M+2, the label could represent one labeled acetyl unit in a fatty-acyl chain, while an M+3 shift may arise from a glucose-derived glycerol backbone. Larger shifts can combine backbone and multiple acyl-chain contributions. For phosphatidylcholine, connected one-carbon and choline metabolism can further complicate interpretation in some tracer settings. Molecular formula, adduct, isotope purity, and the number of potentially labeled atoms must be incorporated into the expected envelope.

MS/MS can strengthen localization by demonstrating label in fatty-acyl fragments or class-specific fragments. However, the absence of a labeled fragment may reflect insufficient product-ion signal rather than absence of label. A targeted assay is valuable after discovery has defined the lipid species, adducts, retention times, and transitions that provide stable label-location evidence. The final identification level should not exceed the structural resolution of the acquired data.

Natural-Abundance Correction and Isotopomer Spectral Analysis

Every unlabeled lipid already contains a distribution of heavier isotopes, principally 13C but also 2H, 15N, 18O, and isotopes of sulfur or other elements where applicable. As carbon number increases, the natural M+1, M+2, and higher peaks become substantial. Raw isotopologue areas therefore overestimate tracer incorporation unless the natural-abundance contribution is removed.

Correction must use the correct elemental formula, ion/adduct composition, charge, tracer isotope purity, and measured mass-isotopologue channels. The formula of [M+H]+ is not interchangeable with [M+NH4]+ or [M-H]−. For intact lipids, an annotation at sum-composition level may still support formula-based correction, but isobaric features and unresolved coelution can distort the measured envelope. Unlabeled matrix controls reveal whether the theoretical correction matches the actual platform and sample background.

High mass resolution does not make correction optional. Orbitrap measurements can show mass-dependent spectral inaccuracy when minor isotopologues are measured alongside a dominant monoisotopic peak. Acquisition settings, detector response, ion statistics, and peak integration should be evaluated with standards or pooled samples across the expected enrichment range. Corrected negative values, non-monotonic patterns, or residual label in unlabeled controls should trigger review of peak boundaries, interference, formula assignment, and correction parameters. Silently replacing all negative corrected values with zero can conceal a method failure and bias low-enrichment estimates.

ISA uses the complete pattern of mass isotopologues rather than a single labeled peak. In the classical fatty-acid model, repeated condensation of two-carbon units creates a binomially related distribution determined by the effective enrichment of the lipogenic acetyl-CoA pool and the fraction of the product synthesized during tracer exposure. The parameters are often represented as D, the fraction of labeled precursor in the lipogenic acetyl-CoA pool, and g(t), the fraction of the measured fatty acid newly synthesized during time t. Extensions can represent elongation of an existing shorter-chain precursor.

Figure 2. The M+0, M+2, M+4, M+6, and higher pattern contains information about effective precursor enrichment and the fraction newly synthesized. Natural-abundance correction and an unlabeled control precede model fitting; a single labeled peak cannot identify both parameters.

ISA is most informative when multiple isotopologues are measured precisely and the product is formed from a defined repeated subunit. A fit can be numerically acceptable yet biologically misleading if several precursor pools contribute, the pool is not homogeneous, labeled product is rapidly lost, or the assumed synthesis route does not match the model. Parameter confidence intervals, residuals, goodness-of-fit criteria, and sensitivity to excluded isotopologues should be retained. A reported g(t) without D, uncertainty, time, product definition, and model description is incomplete.

Distinguishing De Novo Synthesis, Elongation, and Desaturation

Palmitate is the most direct fatty-acid endpoint for mammalian DNL because fatty acid synthase commonly releases a 16-carbon product. Stearate can arise through elongation of palmitate, and palmitoleate or oleate can arise through desaturation of palmitate or stearate, respectively. The isotopologue patterns of these downstream products contain information about precursor-product relationships, but they should not be combined into one DNL value without an explicit model.

Elongation can be evaluated by comparing labeling of a longer-chain fatty acid with its candidate precursor. A pre-existing unlabeled palmitate molecule that receives a labeled two-carbon unit can produce M+2 stearate even when the palmitate backbone was not synthesized during the tracing period. Conversely, fully labeled de novo palmitate may be elongated with an unlabeled unit. Both routes can generate overlapping mass shifts. Time-course data and an elongation-aware ISA model help separate these possibilities.

Desaturation does not add carbon, so a product such as palmitoleate can inherit the carbon isotopologue pattern of palmitate. A change in labeled palmitoleate may reflect altered palmitate synthesis, altered SCD activity, or changes in product removal. Static ratios such as 16:1/16:0 and 18:1/18:0 are sometimes used as desaturation indices, but they are pool-size ratios, not direct desaturation fluxes. Tracer-resolved precursor-product analysis is needed when the mechanistic claim concerns desaturation.

The same distinction applies to complex lipids. Increased labeled TG may result from greater fatty-acid synthesis, more glycerol-backbone labeling, increased esterification of an unchanged fatty-acid pool, or reduced TG turnover. In experimental models exposed to a high-fat diet, DNL and desaturation can be suppressed while elongation and TG synthesis remain active. This illustrates why total lipid accumulation, fatty-acid synthesis, and assembly of storage lipids must remain separate analytical outputs.

Pool Size and Lipid Synthesis Flux Interpretation

Fractional labeling is often inversely influenced by pool size. A small rapidly turning pool can become highly labeled while representing little material; a large pool can show modest enrichment while accumulating a substantial amount of newly synthesized lipid. Quantitative interpretation therefore requires both isotopic and abundance measurements.

Within a qualified model, the newly synthesized amount can be approximated as the fraction newly synthesized multiplied by the measured product pool. The units and denominator must be explicit: nanomoles per million cells, micrograms per gram tissue, or another biological basis. Internal standards and response factors determine whether the pool is reported as absolute amount, class-normalized amount, or relative abundance. Fractional synthesis multiplied by a relative peak area does not become an absolute synthesis rate.

Rate adds time, but division by the labeling interval is valid only when the modeled accumulation is sufficiently linear or the kinetic model accounts for curvature. Product degradation, secretion, re-esterification, and transfer between lipid classes mean that net pool accumulation may differ from gross synthesis. In proliferating cells, dilution by cell division and membrane biogenesis also affects pool behavior. The report should therefore use "apparent synthesis rate," "net newly synthesized amount," or another qualified term when the experiment cannot resolve turnover and export.

Four interpretations should remain visibly separate:

• Increased enrichment with unchanged pool size indicates a larger labeled fraction, but the cause may be greater synthesis or a change in precursor enrichment.
• Increased pool size with unchanged fractional synthesis can still produce more newly synthesized material.
• Decreased enrichment with increased pool size may represent dilution by exogenous lipid uptake rather than pathway inhibition.
• Similar total DNL with different intact-lipid labeling can indicate redistribution of newly synthesized fatty acids among storage and membrane pools.

Figure 3. Corrected mass-isotopologue distributions, time, precursor enrichment or ISA estimates, and quantitative product pools support progressively stronger outputs: tracer contribution, fractional synthesis, newly synthesized amount, and a qualified rate. Separate branches are required for elongation, desaturation, and intact-lipid allocation.

Group comparisons should propagate uncertainty from isotopologue measurement, precursor enrichment, model fitting, and pool quantification. A statistically significant difference in a fitted parameter is credible only if model fit and analytical quality are comparable across groups. When a perturbation changes cell number, viability, water content, or total lipid recovery, normalization should be predefined and accompanied by the unnormalized values needed to understand the denominator.

Reporting Stable-Isotope Lipidomics Results

A stable-isotope lipidomics report should allow the biological conclusion to be reconstructed from the acquired evidence. Summary heatmaps are useful for pattern recognition, but they cannot replace the corrected isotopologue distributions, precursor information, pool measurements, and model diagnostics from which the conclusion was derived.

At minimum, the report should include:

• tracer identity, isotopic purity, labeled fraction, total precursor concentration, exposure interval, and administration or medium-switch procedure;
• biological model, group structure, replicate definition, sampling times, biomass or tissue normalization, and viability or handling covariates;
• extraction timing, internal-standard addition point, analytical platform, ion/adduct definition, chromatographic method, and acquisition mode;
• raw and natural-abundance-corrected M+0 through M+n distributions for the modeled products, with unlabeled-control behavior;
• measured water or intracellular precursor enrichment where available, or the effective precursor-enrichment parameter and assumptions used by ISA;
• product pool size, calibration or response-factor status, and the units used to calculate newly synthesized amount;
• model parameters, confidence intervals, residuals, fit criteria, excluded channels, and rules for values below quantification limits;
• label-localization evidence for intact lipids, including diagnostic MS/MS fragments when the conclusion concerns fatty-acyl rather than backbone incorporation;
• separate results for DNL, elongation, desaturation, and lipid-class allocation, without treating one as a surrogate for another; and
• pooled-QC performance, batch drift, blank findings, missingness, correction exceptions, and sensitivity analyses that could change the biological interpretation.

The language of the conclusion should match the evidence tier. Corrected enrichment supports "incorporation of tracer-derived atoms increased." An ISA result supports "the modeled fraction newly synthesized during the labeling interval increased," provided model fit is adequate. A quantitative pool and time-resolved model may support "the apparent synthesis rate increased." None of these statements independently establishes the activity of a single enzyme or excludes changes in substrate uptake, compartmental dilution, lipid turnover, or secretion.

When discovery data will be transferred to a fixed molecular-species panel, the report should also define the transition criteria for a targeted lipidomics assay, including retained adducts, retention windows, diagnostic fragments, calibration status, and isotopologue channels. This prevents a targeted follow-up from silently changing the structural or quantitative definition established during discovery.

Interpretability depends on alignment among the tracer, molecular readout, correction procedure, kinetic design, and quantitative model. When these elements refer to the same biological quantity, an isotopic pattern can support a reproducible statement about de novo fatty-acid synthesis and the destination of newly synthesized lipid.

References

1. Kharroubi AT, Masterson TM, Aldaghlas TA, Kennedy KA, Kelleher JK. Isotopomer spectral analysis of triglyceride fatty acid synthesis in 3T3-L1 cells. American Journal of Physiology. 1992;263:E667-E675. https://doi.org/10.1152/ajpendo.1992.263.4.E667
2. Lee WN, Bassilian S, Guo Z, et al. Measurement of fractional lipid synthesis using deuterated water (2H2O) and mass isotopomer analysis. American Journal of Physiology. 1994;266:E372-E383. https://doi.org/10.1152/ajpendo.1994.266.3.E372
3. Fu X, Deja S, Fletcher JA, et al. Measurement of lipogenic flux by deuterium resolved mass spectrometry. Nature Communications. 2021;12:3756. https://doi.org/10.1038/s41467-021-23958-4
4. Triebl A, Wenk MR. Analytical Considerations of Stable Isotope Labelling in Lipidomics. Biomolecules. 2018;8(4):151. https://doi.org/10.3390/biom8040151
5. Lane AN, Fan TWM, et al. Isotopomer analysis of lipid biosynthesis by high resolution mass spectrometry and NMR. Analytica Chimica Acta. 2009;651(2):201-208. https://doi.org/10.1016/j.aca.2009.08.032
6. Schlame M, Xu Y, Erdjument-Bromage H, Neubert TA, Ren M. Lipidome-wide 13C flux analysis: a novel tool to estimate the turnover of lipids in organisms and cultures. Journal of Lipid Research. 2020;61(1):95-104. https://doi.org/10.1194/jlr.D119000318
7. Li J, Hoene M, Zhao X, et al. Stable isotope-assisted lipidomics combined with nontargeted isotopomer filtering, a tool to unravel the complex dynamics of lipid metabolism. Analytical Chemistry. 2013;85(9):4651-4657. https://doi.org/10.1021/ac400293y
8. Su X, Lu W, Rabinowitz JD. Metabolite Spectral Accuracy on Orbitraps. Analytical Chemistry. 2017;89(11):5940-5948. https://doi.org/10.1021/acs.analchem.7b00396
9. Duarte JAG, Carvalho F, Pearson M, et al. A high-fat diet suppresses de novo lipogenesis and desaturation but not elongation and triglyceride synthesis in mice. Journal of Lipid Research. 2014;55(12):2541-2553. https://doi.org/10.1194/jlr.M052308
10. Collins JM, Neville MJ, Pinnick KE, et al. De novo lipogenesis in the differentiating human adipocyte can provide all fatty acids necessary for maturation. Journal of Lipid Research. 2011;52(9):1683-1692. https://doi.org/10.1194/jlr.M012195
11. Nasimi H, Holland LKK, Clemmensen KKB, et al. Tracer-assisted shotgun lipidomics (TASL): A quantitative workflow integrating stable-isotope tracing with global lipidome profiling. Analytica Chimica Acta. 2026;1400:345317. https://doi.org/10.1016/j.aca.2026.345317